Project: 11179 (Run 2, Clone 84, Gen 9)

[Qinsp]

This GPU WU is significantly slowing down the operation of a HP-dv8 nVidia 230m chipped machine.

Am I mistaken, or is FAH not supposed to affect the machine operation?

This is a "stupid fast" notebook computer with an i7 and 8gb ram, one of the fastest notebooks made in 2010.

2 weeks ago, FAH had no effect on performance. Now it acts like a netbook.

[RAH]

The GPU client does not release GPU cycles, like the CPU does. Don't think it ever did, and don't think it ever will.

[codysluder]

The GPU client does not release GPU cycles, like the CPU does. Don't think it ever did, and don't think it ever will.

And the GPUs that they put in laptops are designed to be miserly with power consumption so there's not a lot of capacity to fold, and if you do, there's not a lot left for screen refreshes.

Somebody once suggested that if the GPU is the choke point for both FAH and for your OS and you really want to fold, there's a setting in Windows to DISABLE hardware video optimizations. It will use the CPU to build the screen images, which will be slower than if you had an idle GPU just waiting to generate those images, but it may end up being faster than waiting for the next opening in the GPU queue. Video responsiveness should also be more uniform.

I've never done this so I'm not sure if it will help or not, but it's worth a try. Let us know.

[PantherX]

If you mean desktop lagging, then: (From my GPU Guide)
6) Some users complain about desktop lagging when performing basic tasks. Possible solutions are:
A) Enable Aero if you are using Windows Vista or Windows 7.
B) Use this Nvidia Environment Variable; FAH_GPU_IDLE and use a value between 5 and 10 (Guide 1, Guide 2)
Do note that the Environment Variable is for FahCore_11 and its WUs. It doesn't work yet for FahCore_15.

[Qinsp]

Thanks, I will try that.

[toTOW]

Am I mistaken, or is FAH not supposed to affect the machine operation?

That's true for CPU client, but unfortunately not always for GPU ones : GPU doesn't have powerful schedulers like CPU have that allow the system to distribute the load between applications. When you give something to a GPU, it will compute it, even if another task (like UI refresh) need it

[Qinsp]

Found out another interesting point accidentally. With this notebook set on Kill, running a SMP (6701) + GPU (11179), if the power is interrupted (we had a blackout last night), it will instantly shut off the computer. I've never seen that with a notebook before, and apparently is a bug with the hardware/bios. I've run this computer quite a bit on battery running intense applications like 3D realtime games, and while it drains the battery rapidly (40-50 minute life), it has never just died.

What is weird, is that I believe this new behaviour. I started folding about 6 weeks? ago, and now have 6 machines GPU folding. There was no effect on any of the computers until the last two weeks. Now the only computer that "likes" GPU jobs is a nVidia GT220 machine. The rest have become intrusive. No problems with any of the SMP jobs (7 machines SMP, 10 total machines?), other than fan noise, you don't see any effect. I've seen many folk post that they have never seen GPU lag, so it is a bit puzzling how I can see it on 5 out of 6 machines.

In any case, I'll play with environment variables later in the week. It might just be that GPU folding isn't something that's going work in my particular situation.